Claude Sonnet 5 vs Pulse AI
A raw model against a dedicated platform, and the model is twenty-one points ahead. Claude called directly was more accurate on nearly every kind of document and returned nothing empty. Pulse AI returned nothing on several ordinary one-page files. Pulse has a real product, a generous free tier, and an edge on Chinese and Japanese scans. It doesn't publish what a page costs.
Claude here is a general model with a file and a field list and nothing around it. Pulse is a dedicated extraction platform with a pipeline builder, webhooks, and SDKs. In our benchmark the model came second and the platform seventh, which is the kind of result a buyer should take seriously before paying for a product.
Claude led on almost every group: single pages, multi-page statements, long registers, non-English documents, scans. Across the whole test only DocuPipe was ahead of it. Pulse's one clear win was Chinese and Japanese scans, where it was among the better systems and Claude was merely good. Pulse's worst results were on documents that should have been easy, a one-page purchase order, an invoice, a structured export, each returned empty while Claude and everyone else read them near perfectly.
Resilience is where the gap is widest. Claude has nothing to tune and returned nothing empty, with a couple of ordinary forms misread as its worst failures. Pulse's ranked result needed its top model tier with extended reasoning switched on, and its empties fell on files with no pattern, which is the hardest kind of failure to plan around.
On price Claude is a few cents a page on token billing. Pulse gives 20,000 free credits and $0.015 per credit after, but doesn't publish credits per page, so the cost of a page is unknown until you ask, and everything beyond the free plan is custom. We can't tell you which is cheaper in production.
On product Pulse has the real thing. Pulse Platform assembles pipelines from classify, extract, schema, tables, charts, and split steps, saves presets, and shows results with bounding boxes. There are async jobs, batch endpoints, and webhooks on the free tier, SDKs in Python and TypeScript, and a security page listing SOC 2, ISO 27001, HIPAA, and GDPR with self-hosted options on Enterprise. Retention is set up through support rather than in the product. Claude has an excellent general API and no document product at all.
So Claude, unless your corpus is specifically Chinese and Japanese scans and you want a platform rather than a model. In that case, trial Pulse on its free tier with your own files, look at what comes back empty, and get a per-page price before deciding.
Almost any corpus where accuracy matters and you have, or will build, your own pipeline.
Chinese and Japanese material, when you want a platform with a pipeline builder and webhooks and have verified your files come back on the free tier.
How it scores, area by area
The accuracy score comes from the benchmark. Every other score is ours, with the reason written next to it. We use the same weights on every review and comparison.
| Area | Claude Sonnet 5 | Pulse AI |
|---|---|---|
| Accuracy35% of total | 9.2Ranked 2nd of 8 systems on the benchmark. | 7.1Ranked 7th of 8 systems on the benchmark. |
| Resilience15% of total | 7.0No settings to tune and nothing returned empty, but a few ordinary forms came back badly misread with no warning, and there is no layout stage to catch them. | 3.0Ordinary one-page files came back empty with no pattern to predict it, and the ranked result needed the top model tier with extended reasoning turned on. |
| Price15% of total | 9.0$2 and $10 per million tokens; a few cents per page at most. | 5.0$0.015 per credit, but credits per page are not published, so the cost of a page is unknown. |
| Getting started10% of total | 3.0No document product. You build upload, review and results tooling yourself. | 7.0Platform UI with ten seats and 20,000 free credits; everything past that is custom pricing. |
| API and integration10% of total | 5.0Excellent general SDKs, native PDF input, a Files API, batch mode and structured outputs; no document jobs, webhooks or pipeline. | 8.0Webhooks on the free tier, Python and TypeScript SDKs, async and batch endpoints. |
| Data control7.5% of total | 5.0Platform policy only; zero retention for eligible organisations. Nothing document-specific. | 6.0Retention configured by contacting support; zero retention and regional residency on Pro. |
| Compliance7.5% of total | 6.0Platform certifications apply to the API; your pipeline is your responsibility. | 8.0SOC 2, ISO 27001, HIPAA and GDPR listed; self-hosted and VPC on Enterprise. |
| Overallweighted | 7.2weighted score out of 10 | 6.2weighted score out of 10 |
How each area is scored
- Accuracy (35%). The benchmark result, on a 0 to 10 scale.
- Resilience (15%). Does it just work? Fewer modes to choose, no tuning to reach the ranked result, no runs that fail outright, and no documents that come back quietly empty. Informed by the benchmark, judged by us.
- Price (15%). Published cost per page for parse plus extraction, cheapest scoring highest. Unpublished pricing is marked down.
- Getting started (10%). How far you get without a sales call: a dashboard, a free tier with no card, help authoring schemas, a test environment.
- API and integration (10%). Webhooks and which tier they are on, SDKs, no-code connectors, whether the API is built around document jobs at all.
- Data control (7.5%). Deletion controls, scheduled retention, zero-retention availability and on which tier, choice of region.
- Compliance (7.5%). Certifications stated on public pages: SOC 2, ISO 27001, HIPAA with a BAA, GDPR, private deployment options.
By document type
Average accuracy on each group of benchmark documents, both systems on the same files. Every document and score is on the evidence page.
| Document type | Claude Sonnet 5 | Pulse AI |
|---|---|---|
| Statements, filings, reports2 to 5 pages · 21 documents | 92.3% Good | 87.2% Good |
| Short forms, invoices, receiptsone page · 30 documents | 91.7% Good | 72.7% Mixed |
| Long registers, catalogs, directories6 pages and up · 21 documents | 91.3% Good | 52.2% Weak |
| Non-English documents12 languages · 25 documents | 85.2% Good | 65.7% Mixed |
| Scans and photosimages, not PDFs · 7 documents | 80.1% Good | 65.4% Mixed |
Price, product, and policies
Full reviews: Claude Sonnet 5, Pulse AI.
| Claude Sonnet 5 | Pulse AI | |
|---|---|---|
| Pricing model | Per token through the Anthropic API. No document-specific product or plan. | Credits. 20,000 free credits on the free plan, then $0.015 per credit. Credits per page are not published. |
| Entry price | $2 per million input tokens, $10 per million output tokens. | Free plan, then $0.015 per credit. Pro and Enterprise on request; annual billing 15% off. |
| Cost per page | Depends on page size and how much you ask it to return; typically a fraction of a cent to a few cents per page. | Not derivable from the public page, since credits per page are not published. |
| Free tier | No ongoing free tier for API use. | Yes. 20,000 credits. |
| Pay as you go | Yes. Pure usage billing. | Yes, on the free plan via overage. |
| Dashboard | No document product. A console workbench for prompts; upload, review and results tooling is yours to build. | Yes. Pulse Platform: upload files, build pipelines from classify, extract, schema, tables, charts and split steps with saved presets, and see results with bounding boxes. 10 seats on the free plan. |
| Webhooks | None for document jobs. You run the pipeline. | Yes, including on the free plan. |
| API and SDKs | Official SDKs in Python, TypeScript, Java, Go, Ruby, C# and PHP. PDF input up to 600 pages per request, a Files API, batch processing at reduced cost, structured outputs and citations. | Python, TypeScript and REST. Async jobs, batch endpoints and saved pipeline presets. |
| Data retention | Governed by Anthropic's API data policies; zero-retention arrangements exist for eligible organizations. Nothing document-specific. | Retention windows and delete-on-first-poll are configured per organization by contacting support; zero-data-retention on Pro and up; US or EU residency on Pro. |
| Compliance | Anthropic's platform certifications apply to the API; your pipeline around it is your responsibility. | SOC 2, ISO 27001, HIPAA (with agreement), GDPR listed on the security page. Self-hosted and VPC deployment on Enterprise. |
We checked these against the vendors' own pricing and documentation pages on 2026-09-11. Sources: Claude Sonnet 5: platform.claude.com; Pulse AI: www.runpulse.com, docs.runpulse.com. Prices change, so confirm before you budget.
Frequently asked
Is Claude more accurate than Pulse AI?
Yes, by about twenty-one points overall (91.73% to 70.95%). Pulse was ahead only on Chinese and Japanese scans, and returned nothing on several ordinary one-page documents that Claude read correctly.
Is Pulse AI cheaper than Claude?
Unknown. Claude is a few cents a page on token billing. Pulse publishes a per-credit rate but not credits per page, so the cost of a page can't be worked out from public information.
Scores are macro-average field accuracy from the DocuBench 0.2.0, 72-document run. Every document, answer key, and raw extraction is open in the explorer; the scoring rules are on the methodology page, and the replication kit re-runs the scorer. A document a system could not process scores zero rather than being dropped.